ISA DeepData

Extracting occurrence data

Let’s first create a new ID column, this will be used later to link together the measurements and occurrences, and to select records by dataset. We cannot use occurrenceID here because these are not unique within the dataset.

library(uuid)

records$id <- UUIDgenerate(use.time = NA, n = nrow(records))
stopifnot(length(unique(records$id)) == nrow(records))

Now we can select and process the columns that will go into the occurrence table.

extract_occurrences <- function(df) {
    df %>%
      select("id", "dataset_id", "Occurrence", "Event", "Location", "Identification", "Record-level", "Taxon") %>%
      jsonlite::flatten() %>%
      rename_all(~str_replace(., ".*\\.", "")) %>%
      as_tibble()
}

occ <- extract_occurrences(records)

Initial cleanup of occurrence data

First clean up any escaped newlines, empty strings, and placeholder values. Also fix basisOfRecord and convert coordinates to numeric values:

library(stringr)

occ <- occ %>%
  mutate_all(~gsub("\\n", "", .)) %>%
  mutate_all(~na_if(., "")) %>%
  mutate(across(where(is.character), str_squish)) %>%
  mutate_all(~replace(., . %in% c("indet", "Not Reported", "indet."), NA)) %>%
  mutate(basisOfRecord = "HumanObservation") %>%
  mutate(
    decimalLongitude = as.numeric(decimalLongitude),
    decimalLatitude = as.numeric(decimalLatitude)
  )

Let’s check for coordinates issues:

robis::map_ggplot(occ)

Let’s take a look at scientificName and scientificNameID.

occ %>%
  group_by(scientificName) %>%
  summarize(records = n()) %>%
  arrange(desc(records)) %>%
  rmarkdown::paged_table()

occ %>%
  group_by(scientificNameID) %>%
  summarize(records = n()) %>%
  arrange(desc(records)) %>%
  rmarkdown::paged_table()

So at least in the current version at the time of writing (June 2021) there are some quality issues for scientificName.

Fixing taxonomy

Let’s try to clean up the scientific names before we do taxon matching with WoRMS. Here I’m using the gni_parse() function from the taxize package, which connects to the GNI name parser. If a name cannot be parsed, I’m keeping the original.

The first step is to create a list of all distinct names in the taxonomy columns.

taxonomy <- occ %>%
  select(phylum, class, order, family, genus, scientificName)
names <- na.omit(unique(unlist(taxonomy)))

Then pass through the name parser:

library(taxize)

clean_name <- function(name) {
  parsed <- tryCatch({
    res <- gni_parse(name)
    stopifnot(nrow(res) == 1)
    return(res$canonical[1])
  },
  error = function(cond){
    return(name)
  })
}

names_clean <- sapply(names, clean_name)  

Now use the cleaned names for taxon matching:

library(worrms)

match_name <- function(name) {
  lsid <- tryCatch({
    res <- wm_records_names(name)
    matches <- res[[1]] %>%
      filter(match_type == "exact" | match_type == "exact_genus")
    if (nrow(matches) > 1) {
      message(paste0("Multiple exact matches for ", name))
    }
    return(matches$lsid[1])
  }, error = function(cond) {
    return(NA)
  })
}

lsids <- sapply(names_clean, match_name)

Now we need to find the lowest taxonomic level at which we find a name. Note that this will result in records with less taxonomic resolution than intended. Ideally we would only match on scientificName. First translate the taxonomy columns to LSIDs:

taxonomy_clean <- taxonomy %>%
  mutate_all(~names_clean[.]) %>%
  mutate_all(~lsids[.])

taxonomy_clean

## # A tibble: 47,372 × 6
##    phylum         class          order         family genus scientificName      
##    <chr>          <chr>          <chr>         <chr>  <chr> <chr>               
##  1 urn:lsid:mari… urn:lsid:mari… <NA>          <NA>   <NA>  urn:lsid:marinespec…
##  2 urn:lsid:mari… urn:lsid:mari… urn:lsid:mar… <NA>   <NA>  urn:lsid:marinespec…
##  3 urn:lsid:mari… urn:lsid:mari… urn:lsid:mar… <NA>   <NA>  urn:lsid:marinespec…
##  4 urn:lsid:mari… urn:lsid:mari… urn:lsid:mar… <NA>   <NA>  urn:lsid:marinespec…
##  5 urn:lsid:mari… urn:lsid:mari… urn:lsid:mar… <NA>   <NA>  urn:lsid:marinespec…
##  6 urn:lsid:mari… <NA>           <NA>          <NA>   <NA>  urn:lsid:marinespec…
##  7 urn:lsid:mari… urn:lsid:mari… <NA>          <NA>   <NA>  urn:lsid:marinespec…
##  8 urn:lsid:mari… urn:lsid:mari… <NA>          <NA>   <NA>  urn:lsid:marinespec…
##  9 urn:lsid:mari… urn:lsid:mari… <NA>          <NA>   <NA>  urn:lsid:marinespec…
## 10 <NA>           <NA>           <NA>          <NA>   <NA>  <NA>                
## # … with 47,362 more rows

The find the most specific one for each row:

taxonomy_clean <- taxonomy_clean %>%
  mutate(best = coalesce(scientificName, genus, family, order, class))

I’ll use the resulting LSIDs to replace the provided scientificNameIDs.

occ$scientificNameID <- taxonomy_clean$best

Let’s take another look at the top scientificName and scientificNameID after mathing:

occ %>%
  group_by(scientificName, scientificNameID) %>%
  summarize(records = n()) %>%
  arrange(desc(records)) %>%
  head(30) %>%
  knitr::kable()

Extracting MeasurementOrFact data

extract_mof <- function(df) {
    df %>%
      select("id", "dataset_id", "MeasurementOrFact") %>%
      jsonlite::flatten() %>%
      rename_all(~str_replace(., ".*\\.", "")) %>%
      mutate(across(where(is.character), str_squish)) %>%
      mutate_all(~na_if(., "")) %>%
      filter(!is.na(measurementType) & !is.na(measurementValue)) %>%
      as_tibble()
}

mof <- extract_mof(records)
mof

## # A tibble: 12,377 × 6
##    id             dataset_id     measurementID  measurementType measurementValue
##    <chr>          <chr>          <chr>          <chr>           <chr>           
##  1 2d69af50-e297… bgrpmn12015_b… MUC20401639879 Relative abund… 11.9718309859155
##  2 5e86f2d4-bb8f… bgrpmn12015_b… MUC20411639880 Relative abund… 5.63380281690141
##  3 c615232e-467b… bgrpmn12015_b… MUC20421639881 Relative abund… 15.4929577464789
##  4 3238b022-f96c… bgrpmn12015_b… MUC20431639882 Relative abund… 33.0985915492958
##  5 e4f14773-37a4… bgrpmn12015_b… MUC20441639883 Relative abund… 10.5633802816901
##  6 b48d98b6-5827… bgrpmn12015_b… MUC20451639884 Relative abund… 11.9718309859155
##  7 c753b854-512e… bgrpmn12015_b… MUC20461639885 Relative abund… 5.63380281690141
##  8 84afdc08-b9e7… bgrpmn12015_b… MUC20471639886 Relative abund… 0.7042253521126…
##  9 d2d09102-809e… bgrpmn12015_b… MUC20481639887 Relative abund… 0.7042253521126…
## 10 1db232d4-90b2… bgrpmn12015_b… MUC20491639888 Relative abund… 2.8169014084507 
## # … with 12,367 more rows, and 1 more variable: measurementUnit <chr>

A number of records appear to have empty values. To demonstrate this, let’s take a look at the most common combinations of measurementType and measurementValue:

mof %>%
  group_by(measurementType, measurementValue) %>%
  summarize(records = n()) %>%
  arrange(desc(records)) %>%
  head(10) %>%
  knitr::kable()

Generating EML

For demonstration purposes, I’m working with the dataset pertaining to the first record here. The EML template is read from templates/eml.xml:

library(readr)
library(glue)

generate_eml <- function(df) {
  eml <- read_file("templates/eml.xml")
  metadata <- df$Metadata[1,]

  firstname <- strsplit(metadata$Creator$name, " ")[[1]][1]
  lastname <- strsplit(metadata$Creator$name, " ")[[1]][2]
  organization <- metadata$Creator$organisation
  email <- metadata$Creator$email
  position <- metadata$Creator$position

  creator_firstname <- ""
  creator_lastname <- ""
  creator_organization <- metadata$Contact$organisation
  creator_email <- ""
  creator_position <- ""

  abstract <- metadata$abstract
  title <- metadata$title
  citation <- metadata$citation
  packageid <- "https://datasets.obis.org/deepdata"
  pubdate <- format(Sys.time(), "%Y-%m-%d")
  datestamp <- format(Sys.time(), "%Y-%m-%dT%H:%M:%S%z")
  glue(eml)
}

generate_eml(records)

## <eml:eml xmlns:eml="eml://ecoinformatics.org/eml-2.1.1"
##   xmlns:dc="http://purl.org/dc/terms/"
##   xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
##   xsi:schemaLocation="eml://ecoinformatics.org/eml-2.1.1 http://rs.gbif.org/schema/eml-gbif-profile/1.1/eml.xsd"
##   packageId="https://datasets.obis.org/deepdata" system="http://gbif.org" scope="system"
##   xml:lang="eng">
## 
## <dataset>
##   <title xml:lang="eng">BGRPMN12015 BIODIVERSITY</title>
##   <pubDate>2021-11-24</pubDate>
##   <language>eng</language>
##   <abstract>
##     <para>Sampling data captured in Oceanic Exploration Research Based mainly on nodule abundances in 221 box core samples and the hydro-acoustic data obtained during seven cruises to the German license area, resource modelling using artificial neural network statistics was carried out for the entire license area of 75,000 km2 size, and in particular for four prospective potential mining areas in the eastern license area with a total size of 1038 km2. According to these assessments, the entire license area contains 620 million metric tonnes of nodules (dry weight) with 195 million t of Mn, 8.7 million t of Ni, 7.3 million t of Cu, 1.6 million t of Ti, and 1.1 million t of Co. Results
## of geostatistical analyses show that the mean nodule abundance in the four prospective areas in the eastern license area (“sub-clusters”) varies between 17.7 and 23.1 kg/m2 and the resources here amount to 14 million metric tonnes of nodules in total (dry weight), containing 4.4 million t of Mn, 198,000 t of Ni, 165,000 t of Cu, 36,000 t of Ti, and 24,000 t of Co.</para>
##   </abstract>
##   <keywordSet>
##     <keyword>Occurrence</keyword>
##     <keywordThesaurus>GBIF Dataset Type Vocabulary: http://rs.gbif.org/vocabulary/gbif/dataset_type.xml</keywordThesaurus>
##   </keywordSet>
##   <intellectualRights>
##     <para>This work is licensed under a <ulink url="http://creativecommons.org/licenses/by/4.0/legalcode"><citetitle>Creative Commons Attribution (CC-BY) 4.0 License</citetitle></ulink>.</para>
##   </intellectualRights>
##   <maintenance>
##     <description>
##       <para></para>
##     </description>
##     <maintenanceUpdateFrequency>unkown</maintenanceUpdateFrequency>
##   </maintenance>
##   <creator>
##     <individualName>
##     <givenName></givenName>
##     <surName></surName>
##     </individualName>
##     <organizationName>Federal Institute for Geosciences and Natural Resources of Germany</organizationName>
##     <positionName></positionName>
##     <electronicMailAddress></electronicMailAddress>
##   </creator>
##   <metadataProvider>
##     <individualName>
##     <givenName>Sheldon</givenName>
##     <surName>Carter</surName>
##     </individualName>
##     <organizationName>International Seabed Authority</organizationName>
##     <positionName>Database Manager</positionName>
##     <electronicMailAddress>scarter@isa.org.jm</electronicMailAddress>
##   </metadataProvider>
##   <contact>
##     <individualName>
##     <givenName>Sheldon</givenName>
##     <surName>Carter</surName>
##     </individualName>
##     <organizationName>International Seabed Authority</organizationName>
##     <positionName>Database Manager</positionName>
##     <electronicMailAddress>scarter@isa.org.jm</electronicMailAddress>
##   </contact>
## </dataset>
## <additionalMetadata>
## <metadata>
## <gbif>
##   <dateStamp>2021-11-24T21:11:29+0100</dateStamp>
##   <hierarchyLevel>dataset</hierarchyLevel>
##   <citation>Federal Institute for Geosciences and Natural Resources of Germany, (2015). BGRPMN12015 BIODIVERSITY. Available : DeepData, International Seabed Authority https://data.isa.org.jm/ Accessed: [YYYY-MM-DD].</citation>
## </gbif>
## </metadata>
## </additionalMetadata>
## </eml:eml>

Generating an archive descriptor file

The archive also needs to include a meta.xml file which describes the files in the archive and their relationships.

Let’s first get a list of terms including their qualName.

library(xml2)

get_terms <- function(url) {
  doc <- read_xml(url)
  terms <- doc %>%
    xml_ns_strip() %>%
    xml_find_all(".//property") %>% 
    map_df(function(x) {
      list(
        name = xml_attr(x, "name"),
        qual = xml_attr(x, "qualName")
      )
    })
}

occurrence_terms <- get_terms("https://rs.gbif.org/core/dwc_occurrence_2020-07-15.xml")
mof_terms <- get_terms("https://rs.gbif.org/extension/obis/extended_measurement_or_fact.xml")

Using these we can generate a list of terms to go into the meta.xml file for each table.

generate_meta <- function(occ, mof) {
  occurrence_fields <- tibble(name = names(occ)) %>%
    left_join(occurrence_terms, by = "name") %>%
    mutate(index = as.numeric(row.names(.)) - 1)
  
  occurrence_lines <- paste0("<field index=\"", occurrence_fields$index, "\" term=\"", occurrence_fields$qual, "\"/>")
  occurrence_lines[1] <- "<id index=\"0\" />"
  occurrence_lines <- paste0(occurrence_lines, collapse = "\n")

  mof_fields <- tibble(name = names(mof)) %>%
  left_join(mof_terms, by = "name") %>%
  mutate(index = as.numeric(row.names(.)) - 1)

  mof_lines <- paste0("<field index=\"", mof_fields$index, "\" term=\"", mof_fields$qual, "\"/>")
  mof_lines[1] <- "<coreid index=\"0\" />"
  mof_lines <- paste0(mof_lines, collapse = "\n")

  meta <- read_file("templates/meta.xml")
  glue(meta)
}

generate_meta(occ, mof)

## <archive xmlns="http://rs.tdwg.org/dwc/text/" metadata="eml.xml">
##   <core encoding="UTF-8" fieldsTerminatedBy="\t" linesTerminatedBy="\n" fieldsEnclosedBy="" ignoreHeaderLines="1" rowType="http://rs.tdwg.org/dwc/terms/Occurrence">
##     <files>
##       <location>occurrence.txt</location>
##     </files>
##     <id index="0" />
## <field index="1" term="NA"/>
## <field index="2" term="http://rs.tdwg.org/dwc/terms/occurrenceID"/>
## <field index="3" term="http://rs.tdwg.org/dwc/terms/catalogNumber"/>
## <field index="4" term="http://rs.tdwg.org/dwc/terms/recordedBy"/>
## <field index="5" term="http://rs.tdwg.org/dwc/terms/individualCount"/>
## <field index="6" term="http://rs.tdwg.org/dwc/terms/organismQuantity"/>
## <field index="7" term="http://rs.tdwg.org/dwc/terms/organismQuantityType"/>
## <field index="8" term="http://rs.tdwg.org/dwc/terms/sex"/>
## <field index="9" term="http://rs.tdwg.org/dwc/terms/occurrenceStatus"/>
## <field index="10" term="http://rs.tdwg.org/dwc/terms/associatedSequences"/>
## <field index="11" term="http://rs.tdwg.org/dwc/terms/occurrenceRemarks"/>
## <field index="12" term="http://rs.tdwg.org/dwc/terms/eventID"/>
## <field index="13" term="http://rs.tdwg.org/dwc/terms/eventDate"/>
## <field index="14" term="http://rs.tdwg.org/dwc/terms/eventTime"/>
## <field index="15" term="http://rs.tdwg.org/dwc/terms/year"/>
## <field index="16" term="http://rs.tdwg.org/dwc/terms/month"/>
## <field index="17" term="http://rs.tdwg.org/dwc/terms/day"/>
## <field index="18" term="http://rs.tdwg.org/dwc/terms/habitat"/>
## <field index="19" term="http://rs.tdwg.org/dwc/terms/samplingProtocol"/>
## <field index="20" term="http://rs.tdwg.org/dwc/terms/eventRemarks"/>
## <field index="21" term="http://rs.tdwg.org/dwc/terms/locationID"/>
## <field index="22" term="http://rs.tdwg.org/dwc/terms/minimumDepthInMeters"/>
## <field index="23" term="http://rs.tdwg.org/dwc/terms/maximumDepthInMeters"/>
## <field index="24" term="http://rs.tdwg.org/dwc/terms/verbatimDepth"/>
## <field index="25" term="http://rs.tdwg.org/dwc/terms/decimalLatitude"/>
## <field index="26" term="http://rs.tdwg.org/dwc/terms/decimalLongitude"/>
## <field index="27" term="http://rs.tdwg.org/dwc/terms/verbatimCoordinateSystem"/>
## <field index="28" term="http://rs.tdwg.org/dwc/terms/verbatimSRS"/>
## <field index="29" term="http://rs.tdwg.org/dwc/terms/coordinateUncertaintyInMeters"/>
## <field index="30" term="http://rs.tdwg.org/dwc/terms/identificationID"/>
## <field index="31" term="http://rs.tdwg.org/dwc/terms/typeStatus"/>
## <field index="32" term="http://rs.tdwg.org/dwc/terms/identifiedBy"/>
## <field index="33" term="http://rs.tdwg.org/dwc/terms/dateIdentified"/>
## <field index="34" term="http://rs.tdwg.org/dwc/terms/identificationVerificationStatus"/>
## <field index="35" term="http://purl.org/dc/terms/type"/>
## <field index="36" term="http://purl.org/dc/terms/license"/>
## <field index="37" term="http://purl.org/dc/terms/rightsHolder"/>
## <field index="38" term="http://purl.org/dc/terms/accessRights"/>
## <field index="39" term="http://purl.org/dc/terms/bibliographicCitation"/>
## <field index="40" term="http://rs.tdwg.org/dwc/terms/institutionID"/>
## <field index="41" term="http://rs.tdwg.org/dwc/terms/basisOfRecord"/>
## <field index="42" term="http://rs.tdwg.org/dwc/terms/taxonID"/>
## <field index="43" term="http://rs.tdwg.org/dwc/terms/scientificName"/>
## <field index="44" term="http://rs.tdwg.org/dwc/terms/scientificNameID"/>
## <field index="45" term="http://rs.tdwg.org/dwc/terms/kingdom"/>
## <field index="46" term="http://rs.tdwg.org/dwc/terms/phylum"/>
## <field index="47" term="http://rs.tdwg.org/dwc/terms/class"/>
## <field index="48" term="http://rs.tdwg.org/dwc/terms/order"/>
## <field index="49" term="http://rs.tdwg.org/dwc/terms/family"/>
## <field index="50" term="http://rs.tdwg.org/dwc/terms/genus"/>
## <field index="51" term="http://rs.tdwg.org/dwc/terms/taxonRank"/>
## <field index="52" term="http://rs.tdwg.org/dwc/terms/taxonomicStatus"/>
## <field index="53" term="http://rs.tdwg.org/dwc/terms/taxonRemarks"/>
##   </core>
##   <extension encoding="UTF-8" fieldsTerminatedBy="\t" linesTerminatedBy="\n" fieldsEnclosedBy="" ignoreHeaderLines="1" rowType="http://rs.iobis.org/obis/terms/ExtendedMeasurementOrFact">
##     <files>
##       <location>extendedmeasurementorfact.txt</location>
##     </files>
##     <coreid index="0" />
## <field index="1" term="NA"/>
## <field index="2" term="http://rs.tdwg.org/dwc/terms/measurementID"/>
## <field index="3" term="http://rs.tdwg.org/dwc/terms/measurementType"/>
## <field index="4" term="http://rs.tdwg.org/dwc/terms/measurementValue"/>
## <field index="5" term="http://rs.tdwg.org/dwc/terms/measurementUnit"/>
##   </extension>
## </archive>

Bringing it all together

Now we can generate an archive for each dataset. While I’m generating datasets I’m also populating the RSS feed and creating dataset landing pages.

baseurl <- "https://datasets.obis.org/hosted/isa/"
item_template <- read_file("templates/rss_item.xml")
landing_template <- read_file("templates/index_dataset.html")
items <- list()
pubdate <- format(Sys.time(), "%a, %d %b %Y %H:%M:%S %z")

unlink("output", recursive = TRUE)
dir.create("output")

datasetids <- unique(records$dataset_id)

for (datasetid in datasetids) {
  
  dataset <- records %>%
    filter(dataset_id == datasetid) %>%
    head(1)

  dataset$Metadata$abstract <- dataset$Metadata$abstract %>%
    str_replace(., "&", "&amp;") %>%
    str_replace(., ">", "&gt;") %>%
    str_replace(., "<", "&lt;") %>%
    str_replace(., "'", "&apos;") %>%
    str_replace(., "\"", "&quot;")

  title <- dataset$Metadata$title
  abstract <- dataset$Metadata$abstract
  link <- paste0(baseurl, datasetid, "/index.html")
  dwca <- paste0(baseurl, datasetid, "/", datasetid, ".zip")

  # clear dataset directory
    
  unlink(paste0("output/", datasetid), recursive = TRUE)
  dir.create(paste0("output/", datasetid))

  # RSS feed items
  
  item <- glue(item_template)
  items[[datasetid]] <- item
  
  # dataset landing page
  
  landing <- glue(landing_template)
  writeLines(landing, paste0("output/", datasetid, "/index.html"))
  
  # archive  
  
  dataset_occ <- occ %>% filter(dataset_id == datasetid) 
  dataset_mof <- mof %>% filter(dataset_id == datasetid) 

  eml <- generate_eml(dataset)
  meta <- generate_meta(occ, mof)
  
  write.table(dataset_occ, file = paste0("output/", datasetid, "/occurrence.txt"), sep = "\t", row.names = FALSE, na = "", quote = FALSE)
  write.table(dataset_mof, file = paste0("output/", datasetid, "/extendedmeasurementorfact.txt"), sep = "\t", row.names = FALSE, na = "", quote = FALSE)
  writeLines(eml, paste0("output/", datasetid, "/eml.xml"))
  writeLines(meta, paste0("output/", datasetid, "/meta.xml"))
  
  files <- c("occurrence.txt", "extendedmeasurementorfact.txt", "eml.xml", "meta.xml")
  setwd(paste0("output/", datasetid))
  zip(glue("{datasetid}.zip"), files)
  for (f in files) {
    file.remove(f)
  }
  setwd("../..")

}

Generate RSS file

items <- paste0(items, collapse = "\n")
rss_template <- read_file("templates/rss.xml")

title <- "International Seabed Authority (ISA)"
description <- "International Seabed Authority (ISA)"
link <- paste0(baseurl, "index.html")

rss <- glue(rss_template)
writeLines(rss, "output/rss.xml")

Generate landing page

index_template <- read_file("templates/index.html")
content <- paste0(paste0("<li><a href=\"", datasetids, "/index.html\">", datasetids, "</a></li>"), collapse = "\n")
index <- glue(index_template)
writeLines(index, "output/index.html")

Uploading to S3

delete_object("hosted/isa/", bucket = "obis-datasets")
files <- list.files("output", full.names = TRUE, recursive = TRUE, include.dirs = FALSE)

for (file in files) {
  folder <- str_replace(dirname(file), "output", "hosted/isa")
  target <- str_replace(file, "output", "hosted/isa")
  message(target)
  put_object(file, object = target, bucket = "obis-datasets", acl = "public-read")
}